Kernel 3.7 doesn't boot using EFI Stub

adn770 · 2013-07-19 07:30:47

You are right. Seems that the issue I have is a misunderstanding on how rEFInd works as I found that what I'm experiencing is caused by loading the new -custom kernel but picking the initramfs image of the stock kernel which has a different directory name for the modules directory in it and is unable to load them.

Please accept my apologies for disturbing the thread.

srs5694 · 2013-07-19 16:11:27

I just ran across this thread, in which a user reports success at booting a kernel from a USB flash drive but the same kernel fails to boot from a hard disk. This could indicate a timing problem somewhere in the boot loader code. I thought I'd post this as something that's worth testing and perhaps reporting upstream if it pans out. (I'd test it myself, but none of my systems is affected....)

michaels · 2013-07-23 15:56:46

Just for the ones of us who just want a working system (shame on us and credits to those who are working on finally solving this...): Any experience with recent kernels on Lenovo machines? I'm still stuck with 3.9.2...

cfr · 2013-07-23 17:18:13

michaels wrote:

Just for the ones of us who just want a working system (shame on us and credits to those who are working on finally solving this...): Any experience with recent kernels on Lenovo machines? I'm still stuck with 3.9.2...

I think this is just much too generic a question and won't get you much useful info. I have a Lenovo and I think you are lucky 3.9.2 works! Nothing in 3.8.* or later has booted on my machine.

If you just want it to work have you tried using grub or another boot loader?

michaels · 2013-07-23 17:36:51

In the past, I was among that bunch of people with the same working/not-working pattern, so I was looking forward to one of them answering ;-) I'm just a little busy at the moment (especially since Win 7 setup killed the Arch partition on my PC without even asking *ARGH*), so I don't want to test a recent kernel (and rolling back, in case it doesn't work...).

WonderWoofy · 2013-07-23 18:53:10

My machine has been working fine since early in 3.8. I have a Thinkpad E430. There is a "solution" that has worked for some in this thread saying that creating a direct efibootmgr entry causes things to work. You don't have to use that direct entry, but its existence has provided a workaround of sorts for some here. I have had them for quite some time (before this was realized as a workaround), so it may explain why things have consistently worked for me with gummiboot. But in reality, there is no logical explanation why this would make any difference whatsoever. Still, I am hesitant to remove them out of fear that they actually are doing something.

Also, cfr is right. Your question is far too generic for any kind of reasonable response. Even though you were amongst a group of users who were experiencing this on/off situation, as far as I can tell, any given version was not consistently borked or not-borked amongst all users (nor amongst machines of the same model).

If I were in your situation, I would just leave everything set up with gummiboot or rEFInd, and then just use a backup bootloader in the mean time. In fact, I like to keep a couple set up, as this is new technology and I don't yet fully trust it. So I have gummiboot, rEFInd, elilo, grub2, and syslinux-efi all set up. I also have both version of the UEFI shell that are launchable from gummiboot, rEFInd, as well as my firmware boot manager, just in case the "just-in-case backup bootloaders" fail me.

Archcowboy · 2013-08-01 18:17:34

michaels wrote:

Just for the ones of us who just want a working system (shame on us and credits to those who are working on finally solving this...): Any experience with recent kernels on Lenovo machines? I'm still stuck with 3.9.2...

I was brave and just updated from kernel 3.9.2 to kernel 3.10.3 with rEFInd 0.7.1-4 on a Lenovo X230. Boots just fine :)

Last edited by Archcowboy (2013-08-01 18:17:46)

cfr · 2013-08-02 00:23:48

Fails for me. Also on Lenovo (different X121e).

But, honestly, if you just want a working system don't rely on stub loading for the moment. I wouldn't stay on an old kernel just to boot in that particular way. Much too much hassle. As WonderWoofy says, install a backup loader which you can e.g. use from rEFInd's menu. That way, you leave the stub booting in place but have the option of booting without it. (I now default to backup grub from rEFInd and switch to stub loader for testing but either way would work.)

michaels · 2013-08-03 09:09:34

I was brave and just updated from kernel 3.9.2 to kernel 3.10.3 with rEFInd 0.7.1-4 on a Lenovo X230. Boots just fine

Thanks for this information, for me (also X230) it's working too. Still, I installed GRUB as a fallback solution as described here: https://wiki.archlinux.org/index.php/Gr … _systems_2 Because of the vast amount of different information, solutions etc. I didn't think that this was easy, but: it is :-) GRUB is showing up as a menu entry in rEFInd (you don't have to create a efibootmgr entry for it) and can be used if efistub fails.

By the way: Why is it impossible to add an efibootmgr entry from a running system? Do I always have to use CD/USB for this purpose??

Last edited by michaels (2013-08-03 09:10:23)

srs5694 · 2013-08-03 15:34:16

michaels wrote:

Why is it impossible to add an efibootmgr entry from a running system? Do I always have to use CD/USB for this purpose??

You should be able to use efibootmgr from a running system. If you can't, my guess is that you're running into a mismatch of your efibootmgr version and your kernel version -- there have been some recent changes to the kernel's EFI variables support that require matching changes to user-space programs (like efibootmgr). If the two get out of sync, the result is failure of the user-space program. This is just one more EFI headache on top of all the rest.

WonderWoofy · 2013-08-03 21:23:10

I just want to note that Arch does not automatically load efivars anymore. So that is another possibility why you are unable to create those entries. The other possibilities might be that you have a system where >50% of your NVRAM is full, or possibly you might have unknowingly booted via bios compatibility. If it is the case with the too full NVRAM, which was done to avoid bricking systems with crappy NVRAM garbage collection, you can temporarily use the efi_no_storage_paranoia kernel command line parameter. Don't use this all the time though, since if you do have a system with buggy NVRAM garbage collection, you are putting your system at serious risk for bricking.

cfr · 2013-08-03 21:46:41

Interesting. Do you know why it doesn't do that any longer?

Also, update: Matt Fleming asked me to email one of the kernel development lists describing the problem and reporting that commenting the call to setup_efi_pci() allows my kernels to boot. I am not sure what this means. (I mean that I am not sure what the fact that it is being referred to the list means - I know what the analysis of the effect of commenting that call means.)

WonderWoofy · 2013-08-04 02:33:55

Interesting... so you pinned down the issue and determines exactly what was causing your failures?

As far as the efivars, I have no idea why it is not automatically detected (or maybe just loaded) by udev anymore. But I would venture to guess that by not loading the module automatically, it limites access to writing to the nvram. So maybe it is a safety precaution.

cfr · 2013-08-04 02:40:09

I'm not sure. I got rather confused by the fact that for a while the kernel wasn't booting because the kernel wouldn't boot no matter what so I ended up backtracking somewhat and that further taxed my extremely limited understanding of what I was doing in the first place.

It is interesting because /sys/firmware/efi/efivars/ is still populated but perhaps that is read only?

btby · 2013-08-12 04:38:07

The issue is gone for me as of 3.10.5-1 x86_64 on ThinkPad T430, hopefully for good.

Last edited by btby (2013-08-12 04:38:43)

cfr · 2013-08-12 21:23:37

Unsurprisingly, still perfectly consistent for me...

Archcowboy · 2013-08-21 21:52:39

Archcowboy wrote:

michaels wrote:
Just for the ones of us who just want a working system (shame on us and credits to those who are working on finally solving this...): Any experience with recent kernels on Lenovo machines? I'm still stuck with 3.9.2...
I was brave and just updated from kernel 3.9.2 to kernel 3.10.3 with rEFInd 0.7.1-4 on a Lenovo X230. Boots just fine

Updated to 3.10.7-1 and its not working anymore. rEFInd is 0.7.3-1, notebook still the X230.

unikum · 2013-08-22 00:25:21

Archcowboy wrote:

Archcowboy wrote:
michaels wrote:
Just for the ones of us who just want a working system (shame on us and credits to those who are working on finally solving this...): Any experience with recent kernels on Lenovo machines? I'm still stuck with 3.9.2...
I was brave and just updated from kernel 3.9.2 to kernel 3.10.3 with rEFInd 0.7.1-4 on a Lenovo X230. Boots just fine
Updated to 3.10.7-1 and its not working anymore. rEFInd is 0.7.3-1, notebook still the X230.

Same here with 3.10.7-1, gummiboot and Acer V3-771G.

mrunion · 2013-08-22 21:54:57

@unikum: Same laptop as you -- Acer V3-771G-9441. Kernel 3.10.7 just hits the Kernel boot line and freezes.

Gosh, this UEFI stuff is the pits sometimes! I rolled back to 3.10.5 (only rolled the Kernel, not nVidia and everything else) and it works fine again.

mrunion · 2013-08-22 22:46:03

mrunion wrote:

@unikum: Same laptop as you -- Acer V3-771G-9441. Kernel 3.10.7 just hits the Kernel boot line and freezes.
Gosh, this UEFI stuff is the pits sometimes! I rolled back to 3.10.5 (only rolled the Kernel, not nVidia and everything else) and it works fine again.

FWIW: I just did a system update (pacman -Syu) and refind was updated to a .2 release. I let the update install Kernel 3.10.7 and the newest refind (0.7.3-2). I rebooted and still only got to see the kerenl boot line, then the system just halts. No error or nothing. I did find that I can boot to the command line with 3.10.7 and refind, but I can't boot into X. This may make my problem different.

I rolled back ONLY THE KERNEL to 3.10.5 and it boots again both to command line and X.

Does that help anyone?

WonderWoofy · 2013-08-22 22:53:03

@mrunion, on Rod Smith's site (the maintainer of rEFInd) he has a download link for rEFInd compiled against gnu-efi libs instead of tianocore. You should try that and see if it does better with things on your machine. This was mentioned somewhere in this thread, but this thread has started to become a bit unwieldy.

mrunion · 2013-08-22 23:03:11

OK. I may try that. I did have a problem long ago with 3.9.2 not booting, but the next version fixed it and I haven't had issues since. Because of that, I just assumed the tianocore issue didn't affect me.

I've got to work the NASCAR race this weekend here, so I may have a go at that early next week when I recuperate!

WonderWoofy · 2013-08-22 23:10:35

All you actually have to do is download the package, and then drop the refind_x64.efi (and maybe the icons and suff) into place.

cfr · 2013-08-22 23:15:37

@mrunion,
I think you are seeing a different problem and that the alternative binary will not help. If you can boot to the command line, the EFI stub loader is working fine and the problem lies elsewhere. I would guess that the command line you use to boot X is misconfigured somehow. I suggest starting a new thread and posting your config files.

mrunion · 2013-08-22 23:23:45

OK. I think it may be related to the problem with IntelHD cards, etc. that was discussed a few days ago. I will do a bit better research on this.

Arch Linux

#301 2013-07-19 07:30:47

Re: Kernel 3.7 doesn't boot using EFI Stub

#302 2013-07-19 16:11:27

Re: Kernel 3.7 doesn't boot using EFI Stub

#303 2013-07-23 15:56:46

Re: Kernel 3.7 doesn't boot using EFI Stub

#304 2013-07-23 17:18:13

Re: Kernel 3.7 doesn't boot using EFI Stub

#305 2013-07-23 17:36:51

Re: Kernel 3.7 doesn't boot using EFI Stub

#306 2013-07-23 18:53:10

Re: Kernel 3.7 doesn't boot using EFI Stub

#307 2013-08-01 18:17:34

Re: Kernel 3.7 doesn't boot using EFI Stub

#308 2013-08-02 00:23:48

Re: Kernel 3.7 doesn't boot using EFI Stub

#309 2013-08-03 09:09:34

Re: Kernel 3.7 doesn't boot using EFI Stub

#310 2013-08-03 15:34:16

Re: Kernel 3.7 doesn't boot using EFI Stub

#311 2013-08-03 21:23:10

Re: Kernel 3.7 doesn't boot using EFI Stub

#312 2013-08-03 21:46:41

Re: Kernel 3.7 doesn't boot using EFI Stub

#313 2013-08-04 02:33:55

Re: Kernel 3.7 doesn't boot using EFI Stub

#314 2013-08-04 02:40:09

Re: Kernel 3.7 doesn't boot using EFI Stub

#315 2013-08-12 04:38:07

Re: Kernel 3.7 doesn't boot using EFI Stub

#316 2013-08-12 21:23:37

Re: Kernel 3.7 doesn't boot using EFI Stub

#317 2013-08-21 21:52:39

Re: Kernel 3.7 doesn't boot using EFI Stub

#318 2013-08-22 00:25:21

Re: Kernel 3.7 doesn't boot using EFI Stub

#319 2013-08-22 21:54:57

Re: Kernel 3.7 doesn't boot using EFI Stub

#320 2013-08-22 22:46:03

Re: Kernel 3.7 doesn't boot using EFI Stub

#321 2013-08-22 22:53:03

Re: Kernel 3.7 doesn't boot using EFI Stub

#322 2013-08-22 23:03:11

Re: Kernel 3.7 doesn't boot using EFI Stub

#323 2013-08-22 23:10:35

Re: Kernel 3.7 doesn't boot using EFI Stub

#324 2013-08-22 23:15:37

Re: Kernel 3.7 doesn't boot using EFI Stub

#325 2013-08-22 23:23:45

Re: Kernel 3.7 doesn't boot using EFI Stub

Board footer